AI for Skin Disease Detection Only Works Accurately on White Skin Tone: Study

AI for skin disease detection only images datasets does not include vast amounts of skin tones, a new study analysis of the algorithm revealed.

AI Skin Disease Detection

According to the news story by The Guardian, this AI tech is being developed to detect and diagnose skin cancer among other skin-related diseases.

It is to note that artificial intelligence or AI continues to show its potential in further developing healthcare.

Some studies and developments now seek to detect or even to the extent of diagnosing dermatological issues at the same success as health professionals.

It is to note that even Google started developing its own skin condition detection AI last Sept. 16, 2019.

On May 19, the search engine giant revealed during its I/O 2021 event that its AI tool could now help dermatologists to detect skin problems.

However, this method is still in its infancy stage. That said, one of the AI skin disease detection techs turns out to be less accurate for those with a dark skin complexion.

Read Also: Super-Intelligent AI is Hard to Control, According to Researchers | Here's Why it's Beyond Human Comprehension

AI for Skin Disease Detection and Dark Skin Tone

As per the report by The Verge, the majority of the skin tone in this new AI data is from white complexion. On the other hand, darker skin data is only limited in AI.

That said, the study further noted that with the limitation of the AI dataset, the algorithm is likely to only accurately detect skin diseases for those with a white skin tone.

The study from The Lancet Digital Health studied the datasets of the said AI, which includes images of numerous skin conditions, spanning more than 100,000 photos.

However, the analysis further discovered that only 2,236 of the said images contained info regarding the skin color of the person in the photo. On top of that, only 1,400 photos from the expansive data set included the ethnicity of the subject in the photo.

In addition, the study also discovered that the images did not include South Asian, African, and even Afro-Caribbean people.

That said, the study concluded that such a lack of variations and information on the dataset of the AI could pose a high risk of being biased. As such, the AI could end up being available for those featuring a white skin tone.

One of the authors of the study, Dr. David Wen, from the University of Oxford, said that the issue with the AI tech is that it "could lead to certain populations being excluded from algorithms that are approved for clinical use."

The author of the study further said that "the algorithms may not perform as accurately on populations who don't have that many images involved in training."

This article is owned by Tech Times

Written by Teejay Boris